Feeding persistence of mosquitoes (that tendency to attack again within a short period if a first meal is incomplete) is a key determinant of feeding success (defined here as blood-meal volume), especially with regard to mosquitoes that attack defensive hosts. Mortality risk from the vertebrate host is one cost that may be expected to create a trade-off against increased fecundity from second and subsequent meals depending on the availability of blood sources and on the probability of survival until the next gonotrophic cycle. We initially used stochastic simulation models to examine potential trade-offs in fecundity and survival for persistent and nonpersistent lines of mosquitoes that acquire blood from potentially dangerous hosts. In these models, a given strategy, i.e. persistence or the lack thereof, was fixed for the lifetime of the mosquito. Persistent blood feeding paid off relative to nonpersistent mosquitoes when feeding-associated mortality was relatively low to moderate and when the probability of obtaining blood on a given attempt was relatively low. On the other hand, the nonpersistent strategy tended to give relatively higher fecundity when the chances of obtaining blood were high. We also examined the implications of persistence as a feeding strategy for mosquitoes that could exhibit this behaviour in a flexible manner over their lifetime in response to state. We solved for optimal feeding behaviours with a stochastic dynamic programming model set up along similar lines to the simulation model. Results from this approach lend support to the idea that the threshold blood-meal size at which mosquitoes should continue to feed should change relative to the age of the mosquito. The modelling approaches used in this study identified important areas of future study for mosquito blood-feeding behaviour.